I was wondering how Rosetta handles (full atom) chi values when switching from centroid mode to full atom? Does it use some average value? Does it read in the bb angle and look up the dunbrack rotamer value(s)? Or something else?
In the centroid-fullatom switch, there’s a default value of the sidechain dihedrals that is used when the sidechain atoms are first put on.
These values are laughably bad – deliberately so. After the initial atom placement, the sidechains are typically repacked. That is, the rotamers are combinatorially optimized by Metropolis Monte Carlo Simulated Annealing. Having really bad initial values for the chis makes it easier to see when this step gets accidentally omitted.